<p>As one of the effective techniques of image segmentation, fuzzy C-means clustering (FCM) algorithm can perform well in image segmentation for noiseless images, but it is sensitive to outliers. In order to overcome this shortcoming, scholars have proposed various improved methods. The previous ideas of fuzzy clustering are to initialize the membership, and then calculate the cluster center further. Due to the randomness of initialization, the objective functions of some algorithms may oscillate or become larger before converging during the first few iterations, which affects the segmentation speed ultimately. In order to solve this problem, we proposed the cluster center initialization with fast convergence based on FCM (FCM-CCI). Firstly, the pixels of image are evenly divided into groups equal to the number of clusters according to the histogram statistics of the image. Then, the algorithm uses each group of pixels and its corresponding number to calculate the value of the corresponding central pixel, which is the initial clustering of the image. In addition, this algorithm combines the pixels of the original image and the filtered image on the basis of the original FCM algorithm and takes the local density function as the correction weight. Finally, the final segmentation result is obtained by the membership filtering idea. The experiments of performance evaluation on synthetic and real images show that the algorithm has better effect of segmentation and faster convergence. In addition, the convergence speed of some FCM algorithms is further improved by integrating the proposed algorithm into these algorithms.</p>